cs.CL · 2026-11-10 · No. 169
Computation and Language, 2026-11-10.
14 new papers in cs.CL. Titles, authors,
abstracts. Links to arXiv. Want this in your inbox every morning? Subscribe →
01 — The papers
14 entries-
01
HiVe-Mem: Interleaving Horizontal Memory Expansion and Vertical Evidence Composition for Multimodal Agents
Haozhen Zhang, Bohan Liu, Jianzhu Bao, Qingyuan Liu, Haodong Yue, Xiao Li, Zi Pong Lim, Wenya Wang
cs.CL · cs.AI · cs.LG
As multimodal agents accumulate increasingly long interaction histories, effectively retrieving and utilizing past information becomes critical for long-term interaction. Existing memory systems typically rely on semantic retrieval over compressed memories and expose predefined representations, which may miss relevant memories and overlook heterogeneous evidence needs across queries. We introduce \textbf{\hivemem}, a multimodal memory...
-
02
Is Persona Generalization Predictable?
Olina Mukherjee, Arnav Goel, Aditi Raghunathan, Jacob Mitchell Springer
cs.CL · cs.LG
Persona generalization occurs when fine-tuning a language model in one domain induces behavioral changes in others -- for example, the broadly unsafe behavior observed in emergent misalignment. We discover a surprising "on-the-line" phenomenon in persona generalization: in-domain and out-of-domain persona rates across various fine-tuning hyperparameters follow an approximately probit-linear relationship, and the specific probit-linear...
-
03
PORT: Preference Optimization via Robust Token-Level Reweighting
Ding Zhu, Xiukun Wei, Tian Xie, Zhihui Zhu, Xueru Zhang, Mohammad Mahdi Khalili
cs.CL · cs.AI
Preference optimization has become a central approach for aligning large language models with human values, but its effectiveness depends heavily on the quality of preference annotations. In practice, preference data is often noisy due to annotation errors and ambiguity. Existing robust preference optimization methods primarily operate at the sequence level, implicitly treating entire responses as uniformly correct or incorrect. However,...
-
04
The Narrative Gap: Can LLMs Help Us Navigate Diverse Narratives Across Languages?
Nikhil Sharma, Kelly Marchisio, Kenton Murray, Ziang Xiao
cs.CL · cs.AI · cs.HC · cs.IR
Large language models (LLMs) increasingly mediate how people seek and consume information, including accounts that differ across languages and regions, especially on contested topics. Supporting diverse information seeking requires models to help users navigate competing narratives across languages and information streams. We define six capabilities necessary for diverse multilingual information seeking: three foundational capabilities for...
-
05
How Far Does Second-Order Linear Response Predict Quantization Mismatch? Pre-Registered Measurements on Five Language Models
Dezhen Lu, Zixiang Ni
cs.CL · cs.CR · cs.LG
Serving systems choose a weight format for each model and deployment. A low-cost screen predicts the resulting KL divergence from the unquantized model's Fisher information using $\tfrac12Δ^\top FΔ$, the expansion underlying Hessian-aware and second-order post-training quantizers. We assess the reliability of this screen in GPU batches with channel-wise int8, per-tensor fp8 and group-wise int4 on Qwen3-8B, Qwen3-4B, OLMo-2-7B, Qwen3-32B and...
-
06
Adapting Small Language Models to Romanized Nepali: Tokenizer Fertility, Reference Likelihood, and Generation Metrics
Pujan Pant, Sagar Newpane, Hari Sitaula, Manish Sakhakarmy
cs.CL · cs.AI
Romanized Nepali, Nepali written in Latin script, is common in digital communication but has no standard spelling and few resources, and it is unclear how reliably small open language models can be adapted to it. We adapt three 2-3B base models from different families, Llama-3.2-3B, Qwen2.5-3B, and Gemma-2-2B, with QLoRA on 9,000 instruction examples produced by deterministic transliteration of Nepali instruction data. Each model is trained...
-
07
Unmasking Four Knapsack Problems in the Traveling Thief Benchmark
Mohammed Mizan Alanazi, Mahdi Khemakhem
cs.CL · cs.AI
The traveling thief problem (TTP) asks a thief to choose a tour and a packing together, and the benchmark of 9,720 instances of Polyakovskiy et al. (2014) is the standard test set for comparing TTP heuristics. We show that four of these instances are not TTP instances at all. Their files give a renting rate of 0.00, so the tour has no effect on the objective and each instance reduces to a 0-1 knapsack problem. One of them belongs to the 60...
-
08
AgentOPSD: Turn-Level Credit Assignment via Recursive Belief Revision
Yizhuo Li
cs.CL · cs.AI
Reinforcement learning with verifiable rewards ordinarily spreads a single advantage over a whole trajectory, a granularity too coarse to isolate the few decisions on which success in a long, multi-turn interaction actually turns. Privileged self-distillation supplies denser supervision, but it leaves unanswered how a set of local teacher-student discrepancies should be converted into credit along a sequence. We introduce AgentOPSD, a...
-
09
Text Grouping Changes Completion-Time Comparisons in Long-Form Simultaneous Speech Translation
Hieu Minh Truong, Dung D. Le
cs.CL · cs.AI
Does grouping translation updates before speech synthesis change the measured completion gap between two translators? We compare StreamAtt with its fixed-audio-history baseline on ten English-German talks. For each translation, we keep the intended text and its availability times fixed and vary how updates are grouped for synthesis. Four-second grouping increases StreamAtt's mean completion lead from 106.6 to 229.1 seconds with XTTS-v2, but...
-
10
Phrase-Level Speculative Decoding on a Frozen Target
Madhusudhana Naidu
cs.CL · cs.AI · cs.LG
Speculative decoders raise throughput by banking more target tokens per forward pass, and Medusa, EAGLE-3, DFlash and DSpark all do it the same way: more candidates that survive verification. All of them draft in the target's own vocabulary, so one node of the candidate tree is one target token, and the tokenizer makes that unit look like a constant. It is not. BPE's pretokenizer splits on whitespace before merges are learned, so no...
-
11
Prune, Don't Merge: Expert Compression of Mixture-of-Experts LLMs on Multilingual Production Tasks
Tomas Rasymas, Povilas Kvedaras
cs.CL · cs.LG
Mixture-of-Experts language models are cheap to run per token but expensive to hold in GPU memory, which drives the cost of serving them in production. One-shot expert compression shrinks them after a single calibration pass. Its two current state-of-the-art methods are REAP for expert pruning, which deletes the least salient experts, and REAM for expert merging, which merges them into the most salient ones. Both have been evaluated only on...
-
12
TrajLift: Lifting Memory Recall Trajectories with Latency-Modulated Heat Diffusion
Jiawen Kang, Jinchao Li, Mingyu Cui, Junan Li, Xixin Wu, Helen Meng
cs.CL · cs.AI · q-bio.NC
Verbal memory retrieval carries rich diagnostic information for Alzheimer's disease through the order, hesitation, and categorical organization of recalled words, yet computational modeling of this retrieval process remains largely unexplored. A key challenge is that timing and hierarchical structure are diagnostically entangled: the diagnostic meaning of retrieval timing depends on its hierarchical context. We propose TrajLift, a spectral...
-
13
Lost in Rounding: bf16 Recurrent State Makes Hybrid Reasoning Models Think Longer
Yuchuan Weng, Xin Chen, Chi Liu
cs.CL · cs.AI
Hybrid reasoning models interleave attention with recurrent mixers such as Mamba-2 and Gated DeltaNet, whose state is cached between decoding steps. Prior work showed that storing this state in FP16 or INT8 with round-to-nearest (RTN) makes Mamba-2 and GDN hybrids more verbose, and that stochastic rounding (SR) fixes it. We study the other 16-bit format that serving engines expose, bf16, on six open hybrid reasoning models and 590...
-
14
EmoHarness: Boosting Multimodal Emotion Understanding with Meta-Harness Evolving
Fenghai Li, Bingxuan Li, Jiawen Zhang, Yifan Shen, Jiateng Liu, Yutong Zhang, Yue Guo
cs.CL · cs.AI · cs.CV
Small multimodal models can misjudge emotion when words, voice, and visual cues conflict. We present EmoHarness, which runs such a model, the frozen student, inside a harness: the prompts, settings, and verification code around its calls. A human-written starting harness, the scaffold, numbers the student's visual observations, transcript, and prosody descriptions as citable evidence and has the student weigh candidate emotions against it. In...
This edition is part of The Daily Abstract — cs.CL archive. Subscribe to receive these in your inbox each morning, automatically translated to Spanish, with reply-to-PDF: arxivdaily.ignorelist.com.
#D99C5E. Built and served on an always-free VM. The masthead is set 14% letterspaced because newspapers do that and it works.